I seem to have one bad page (second page on first block) on my nand in my JuiceBoxUMDCart.
One of the disadvantages of non-removable media ;-(
Does the ROM bootloader know how to handle bad blocks ?
It might have a mechanism for skipping/remapping bad blocks but the implementation is unknown - AFAIK bad block flagging is usually done by writing a value to the "extra" bytes at the end of a specific page within the block i.e. it is not done by the chip it self, the controlling program does it.
The boot process only needs the first 0x250 bytes or so of NAND to be valid so if the failing bytes are after that point in the first block you may be able to have your program start in the next block.
Have you read back the data to see if it is bad? You may try adjusting the program that is doing the writing, I had a timing issue with a 32M part that turned out to be caused by sending 4 address bytes as opposed to 3. Both 3 and 4 byte addresses are valid for a 32M part but this chip didn't like seeing the 4th byte (probably related to the slow speed that JTAG is switching the control signals). Doesn't sound like what you are seeing but it does show that minor issues can cause problems.
Good luck. 
-J
